Heart Rhythm Classification from Electrocardiogram Signals Using Hybrid PSO-Neural Network Method and Neural ICA

Studies on the classification of heart rhythms from Electrocardiogram (ECG) signal interpretation have been widely reported. Several techniques for recognizing the abnormalities on left bundle branch (LBBB), right bundle branch (RBBB) and premature ventricular contraction (PVC) using the Taguchi optimization method and the Naïve Bayes classification method have been reported. Unfortunately results from the Naïve Bayes classification method are not as good as those using method such as SVM classification method. In the paper we propose a Hybrid PSO-Neural Network (NN) as a classification method and a Neural Independent Component Analysis (Neural-ICA) as a filter method. Neural ICA aims to separate the original signal and the noise signal on the ECG signal record. In this research the ICA method implements the Neural algorithm for the process of updating the weights after filter process. The Hybrid PSO-Neural Network is a Neural Network method that optimized by PSO to optimize the classification result. Hybrid PSO-NN method can improve the classification accuracy up to 2%, i.e. 99% accuracy, in comparison to NN method 98% accuracy and SVM method 96% accuracy, respectively.